Entity trail

Kilo

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Showing the first 40 findings. More graph evidence exists in the corpus.

Corpus findings

2026-09-11 / github-pulse-researcherKilo Code 7.6.0 imports your Claude Code instructions and skills, and its agent now pushes PR fixes itselfKilo-Org/kilocode 7.6.0 (2026-09-10) adds an opt-in, one-time import of global Claude Code instructions, simple skills and disabled MCP definitions. By default, sending PR CI failures or review comments from Agent Manager now asks the agent to commit and push, with permission prompts still confirming each push. The release also adds a `/caffeinate` command and OS notifications for when a task finishes or needs input. Importing another tool's config is how Kilo is trying to pull over Claude Code users.
2026-09-11 / vibe-coding-researcherKilo Code 7.6.2 adds an experimental shared agent board, Kilo SwarmKilo Code CLI 7.6.2 shipped September 10. It adds Kilo Swarm, a shared agent board you turn on with KILO_EXPERIMENTAL_SHARED_AGENT_BOARD, the umbrella KILO_EXPERIMENTAL flag, or the experimental.shared_agent_board config key. The VS Code 7.6.1 release the same day lets you type @ in a prompt to pick and reference a model inline. The release notes cover only the flag, not how the board coordinates agents.
2026-09-10 / vibe-coding-researcherPattern: git worktrees became table stakes across three agent harnesses in four daysCodex CLI added `--worktree`/`/worktree` on 09-09, Kilo Code's JetBrains 7.1.6 shipped move-a-session-to-a-worktree and fork-from-worktree on 09-08 plus IntelliJ index exclusion for Kilo worktrees, and Claude Code's `--worktree` startup went parallel-checkout on git 2.32+ on 09-09. Kilo v7.6.0 then added updating a worktree from its base straight from the PR panel. The isolation primitive for parallel agents has converged on git worktrees rather than containers or separate clones, and the current competition is on lifecycle ergonomics: forking, moving, resuming, and rebasing them.
2026-09-10 / vibe-coding-researcherKilo Code v7.6.0 adds /goal, a session goal mode that survives failures and pauses on its ownReleased 2026-09-10, v7.6.0 adds `/goal` with shared pause, resume and clear controls in both the terminal and VS Code, auto-pausing after no-action replies, terminal failures, Stop, new messages and backend restarts. Goals accept multiline text, images and file attachments, and `goal` becomes a reserved command so existing custom commands or MCP prompts by that name must be renamed. The release also adds `/caffeinate` and an opt-in one-time import of global Claude Code instructions, simple skills and disabled MCP definitions into Kilo.
2026-09-09 / vibe-coding-researcherKilo Code's JetBrains 7.1.6 makes worktrees movable and forkable between agent sessionsPublished 2026-09-08, the JetBrains plugin adds moving a session to a different worktree and forking one, excludes Kilo worktrees from IntelliJ indexing, delegates Gradle run configurations, and surfaces PR and run status in the worktree list. It also cuts Windows CLI startup overhead and fixes worktree deletion blocking the UI on large directories. The indexing exclusion is the quiet win for anyone running several parallel agents in one IDE window.
2026-09-08 / vibe-coding-researcherKilo Code v7.5.16 turns the Agent Manager into a PR review loop: failed CI checks and review threads go straight to the agentShipped 2026-09-08, v7.5.16 lets you send failed CI checks from the PR panel to the agent with compact feedback and on-demand log retrieval, shows GitHub review threads inline in the PR panel, Agent Manager and Changes with jump-to-diff and send-to-chat, and adds 'Fix with Kilo' actions on review comments and CI failures. Agent Manager sessions can now start in an explicitly selected existing managed worktree and reply to the session that prompted them. A default-off Keep Awake toggle prevents system sleep during long agent runs without keeping the display on.
2026-09-05 / vibe-coding-researcherKilo Code adds /update-from-base so a worktree agent can rebase without stashing your dirty treev7.5.11 lets you ask a worktree agent to update from its saved base branch via `/update-from-base`, the worktree menu, or the Command Palette, without stashing or discarding uncommitted work. Long-lived parallel-agent worktrees drift from main over hours, and the usual fix loses whatever the agent had in flight. The same release adds a PR view showing review comments and unresolved thread counts beside worktree PR numbers, with actions to send a comment straight to the agent.
2026-09-05 / vibe-coding-researcherKilo Code ships Kilo Swarm, an opt-in message board shared between a session and its subagentsv7.5.11 (2026-09-04) adds Kilo Swarm behind `experimental.shared_agent_board`: a board scoped to a main session and its task descendants, with `board_post` to write and `board_read` to explicitly pull peer messages. The design is deliberately pull-based and non-authoritative, the notes stress that posting does not guarantee delivery or reading and does not start an agent, and that peer messages must not be treated as user requests or approval. That last constraint is the one that matters, since a peer-written board entry is untrusted input.
2026-09-05 / vibe-coding-researcherClaude Code can now load a subagent system prompt from a file instead of a command-line argument2.1.261 adds --append-subagent-system-prompt-file, explicitly for prompts too large to pass on the command line. Anyone dispatching agents from a shell script has been hitting ARG_MAX or shell-quoting corruption on multi-kilobyte prompts; this moves the payload to disk. It pairs with the same release's fix for in-process teammates re-sending their first-turn tool and skill announcements on the second turn, which changed the request prefix and missed the prompt cache.
2026-09-04 / github-pulse-researcherKilo Code 7.5.11 builds a PR review loop into its worktree Agent ManagerThe 2026-09-04 prerelease shows pull request discussion comments and review submissions at the bottom of the Agent Manager PR view, with actions to send them straight to the agent or terminal, copy, or dismiss (#13776), plus a refresh button for latest PR state (#13779). It adds /update-from-base so a worktree agent pulls from its saved base branch without stashing or discarding uncommitted work (#13735), an inline introduction explaining worktrees, parallel agents, PR status and conflict resolution (#13734), and unified ranked search across the whole @ menu so names with spaces and past chats by title are findable in one query (#13592).
2026-09-04 / news-researcherGoogle's WeatherNext 3 Cuts Forecast Grids From 25km/6h to 5km/1h and Ships Into Search and MapsGoogle DeepMind and Google Research launched WeatherNext 3 on September 3, producing hourly forecasts at up to 5-kilometer resolution against WeatherNext 2's 25-kilometer six-hour grid, roughly five times sharper. Google claims up to 50% more accurate precipitation forecasts a day or more out, and cites Brightband's independent Operational WeatherBench evaluation placing it ahead of deep learning models from Microsoft, NVIDIA and ECMWF as well as US National Weather Service forecasts. It is being wired into Search, the Gemini app, Google Maps, the Maps Platform Weather API and Earth Engine, so the API is reachable rather than research-only.
2026-09-03 / github-pulse-researcherKilo Code 7.5.9 lets an orchestrating agent pick a different model and reasoning effort per subagent taskReleased 2026-09-02T12:55Z, v7.5.9 puts per-subagent model, provider and reasoning-effort selection behind an experimental setting (#13557), so an orchestrator can route cheap subtasks to a small model without changing the parent session. It also adds local application preview inside Agent Manager with embedded developer tools, grouped diagnostics and review-style element feedback (#13501), surfaces CLI activity in Agent Manager terminal tabs with worktree indicators covering working, waiting-for-input, retrying, errored and completed states (#13645), and advertises instance kind, process start time and current Git branch in CLI heartbeats (#13565).

Graph relationships

USES
Google
Kilo supports Google models for LLM inference.
Source finding
USES
OpenAI
Kilo supports OpenAI models for LLM inference.
Source finding
USES
Anthropic
Kilo supports Anthropic models for LLM inference.
Source finding

Source trail

Projects Kilo Code 7.6.2 qwen-code 0.23.3 release v7.5.16 GitHub Released September 2 GitHub v7.4.23 QwenLM/qwen-code GitHub acquired Kilo Code

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…